Exploring the Potential of Quantum Enhanced Multi-Object Tracking and Re-Identification

Multi-object tracking is important for modern precision livestock farming as it enables simultaneous tracking of multiple animals in complex environments (Nidhi et al., 2025). The resulting tracklets of animals can be used to continuously assess the health and welfare of the animals to use for animal management and phenotyping in animal breeding programs. In general, multi-object tracking is a computer vision task that aims to detect and track all objects of interest in a scene, be it people, tomatoes, or animals, with a consistent, unique identifier. In precision livestock farming, multi-object tracking is often approached with tracking-by-detection (Nidhi et al., 2025). Tracking-by-detection uses two algorithms, one to detect the objects of interest, such as You Only Look Once (YOLO, Jocher et al. (2023)) and one to track the detected objects of interest, such as BoT-SORT (Aharon et al., 2022). However, assigning a consistent identifier to each track is difficult due to occlusions that can occur in large groups of animals. As a result, detections can be lost and tracks may be re-initialized, or identities switched between animals. Re-identification (reID) aims to recover the lost or switched identities to maintain a consistent identity for each track. Today, most computer vision tasks run on GPU hardware. Many of these tasks, particularly identity association in multi-object tracking across multiple frames, can be formulated as combinatorial optimization problems (Cornuejols et al., 1977), which may be NP-hard in general. In this context, it is worth exploring whether quantum algorithms can deliver significant speedups over current methods (Meli et al. 2025). Quantum computing is an emerging technology that differs from traditional computing by using quantum bits (qubits) instead of bits on traditional computing hardware. Qubits reflect pieces of quantum information that may exhibit two intrinsic quantum properties: superposition, allowing them to exist in multiple states simultaneously, and entanglement, correlating qubits beyond whatis possible on conventional computers. This unique storage structure allows quantum systems to process information in ways that have no classical counterpart, enabling specialized quantum algorithms. Quantum computers are not intrinsically faster at performing the same operations, but can potentially solve the same computational problems in fewer operations – for example, Shor’s algorithm (Shor, 1994) to factor large integers exponentially faster than any classical methods, rendering current encryption schemes vulnerable once sufficiently large quantum hardware becomes available. Potential applications in life science have recently been proposed by Pook et al. (2025); however, only as a review without developing a concrete quantum algorithm. In this work, we propose a novel extension of the BoT-SORT framework to improve its integrated reID by an additional post-processing step that is based on pairwise comparison of tracked objects at consecutive time points based on the overlap between bounding boxes. Alongside an implementation on traditional hardware, we will also outline the required steps totranslate our developedoptimization problem for execution on quantum hardware and assess the overall suitability of quantum computing for livestock breeding applications.